Using Environmental Conditions to Understand VAST Predictions

Model output from VAST models are inherently hard to digest, particularly when they go against prior held beliefs or intuitions about species poreferences.

The following figures will walk through what predictions have been made for a small subset of species. For each species we will show: 1. Predicted biomass distributions
2. Environmental state at that point in time
3. Species environmental covariates
4. Where each ecological production unit on average falls on those preference curves

Loading the Pieces

1. Predicted Biomass Distributions

Vast model predictions have been processed to return values for the baseline period (2010-2019) as well as at points in the two SSP scenarios when the average SST for the region crosses 0.5, 1, 2, 3, & 4 degrees celsius

These have been processed in: repositories/COCA_Shiny/Coca_SDM_app_dev/spatial_data_mgmt/vast_projection_preprocessing

And can be found in: /Users/akemberling/Documents/Repositories/COCA_Shiny/Data/projections/

2. Environmental State from SSP Scenarios

Average conditions from the SSP scenarios have been processed for mapping conditions during the baseline period (2010-2019) as well as at points in the two SSP scenarios when the average SST for the region crosses 0.5, 1, 2, 3, & 4 degrees celsius.

These environmental states were processed in: temperature_milestone_conditions.ipynb

And can be found in: RES_Data/SSPX_XX/SST_horizon_conditions/

3. Species Preference Curves

The individual preference curves for species have been processed independently and can be found here: Mills_Lab/Projects/COCA19_projections/tables/

These curves need to be transformed back to absolute temperature and depth units in celsius or meters for each of the covariate curves. This is done with the grand mean values from the environmental covariate inputs to the VAST training data.

Grand mean/sd of the covariates can be used to un-scale these relationships.
Depth mean/sd = 123.47/100.2
SST mean/sd = 11.11/4.55
BT mean/sd = 7/2.72

4. Average Conditions Acrosss Ecological Production Units

For Each of these species preference curves we want to overlay what the average temperatures are for an EPU at a certain point in time. The EPU’s have all been masked over the SSP ensembles already, so we just need to get those timeseries and pull out the points in time where these thresholds are crossed.

Timeseries of SSP scenarios are produced in: helper_01_regional_timeseries_construction.R

Files can be found: ``